EU AI Act Omnibus resets high-risk compliance deadlines to 2027-2028
Tags AI / ML · Industry
The EU's AI Act Omnibus, in force since July 27, 2026, pushes the heaviest high-risk AI obligations to Dec 2, 2027 for Annex III systems and Aug 2, 2028 for Annex I embedded AI, while keeping August 2026 transparency duties in place. The amending regulation was adopted by the Council on June 29, 2026 and entered into force three days after Official Journal publication. Article 50 transparency duties and general-purpose AI model rules (Articles 51-55) still applied from Aug 2, 2026, and pre-existing generative systems must watermark output by Dec 2, 2026. New prohibitions on AI generating non-consensual intimate imagery and child sexual abuse material took effect immediately.
Technical significance
Delaying the high-risk obligations to 2027-2028 gives compliance teams in recruitment, credit scoring, education and embedded-AI products more runway, but it does not remove the requirements — it shifts the deadline. The immediate effect is a split regime: transparency and GPAI rules bite now, while the heaviest conformity assessments wait. The new NCII and CSAM prohibitions taking effect immediately signal the EU is unwilling to delay content-harm rules even as it defers procedural ones, and a separate Digital Omnibus covering GDPR, NIS2 and the Data Act is still moving.